Joint TRFI and Deep Learning for Vehicular Channel Estimation

Abdul Karim Gizzini (1) , Marwa Chafii (1) , Ahmad Nimr (2) , Gerhard Fettweis (2)

1 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes
2 Technische Universität Dresden, Vodafone Chair Mobile Communications Systems

IEEE 802.11p standard enables the wireless technology that defines vehicular communications. However, IEEE 802.11p frame structure employing low pilot density is not enough to track the channel variations in high mobility scenarios, leading to significant performance degradation. Therefore, ensuring communication reliability in vehicular environments is considered as a major challenge. In this work, this challenge is tackled by employing deep learning into conventional channel estimation through utilizing deep neural networks (DNN) as an additional non-linear processing unit to correct the interpolation error of the time domain reliable test frequency domain interpolation (TRFI) channel estimates, besides learning higher order statistics of the estimated channel, resulting in a better channel tracking over time. Simulation results demonstrate the performance superiority of the proposed TRFI-DNN scheme over conventional schemes and the recently proposed DNN estimators with a significant computational complexity decrease, especially in high mobility vehicular scenarios.

Mots clés: Channel estimation; deep learning; DNN; IEEE 80211p standard; vehicular communications

Domaines: Traitement du signal et de l'image [eess.SP]; Intelligence artificielle [cs.AI]; Complexité [cs.CC]

Abdul Karim Gizzini, Marwa Chafii, Ahmad Nimr, Gerhard Fettweis. Joint TRFI and Deep Learning for Vehicular Channel Estimation. IEEE GLOBECOM 2020, Dec 2020, Taipei, Taiwan. ⟨hal-02951636⟩